PhD proposal – Data processing from multiple sensors and collaboration of mobile agents for environment exploration
Supervisors: Pr. Sylvain Contassot-Vivier and Dr. Laurent Ciarletta – University of Lorraine, Loria
Collaborative exploration, embedded systems, parallel computing, communication, 3D cartography.
A Master’s degree in computer science or an equivalent diploma is required. Fundamental knowledge in embedded systems as well as good skills in C/C++ programming are required. Additional knowledge in parallel algorithmic will be appreciated.
Context and problem:
We consider a set of mobile elements (agents) with multiple sensors and actuators, that interact with their environment in a coordinated way, not necessarily centralized, to perform a given task. During the last few years, advances in hardware of such systems have induced new possibilities. Indeed, the constantly growing integration of processors implies an increase of embedded computation resources (Nvidia Jetson TX1, parallela card,…) as well as more sophisticated sensors (Kinect, Intel realsense, HD stereo camera Zed, Lidar 3D,…). Yet, as soon as the task is complex enough, it is often necessary to process and merge data from different sensors . However, contrary to the classical computing context, mobile agents require to take into account several important constraints such as the limited computing resource, the communication costs (time and energy) between agents or with external resources, the robustness of the system, and bounded execution times for some tasks (reactivity and safety of the system from a real-time point of view). Thus, in some way we are in a similar case as in , but with yet more constraints. So, it is necessary to design and implement specific algorithmic schemes for processing and merging data that are adapted to those constraints. They must be efficient while having a low energy cost and ensuring some robustness in data exchanges and global coordination .
This PhD is part of a wider project which aims at exploring and calculating accurate 3D reconstructions of environments. So, it is complementary to other works in progress that are focused on the computing resources management. This PhD will address the data processing and merging as well as the planification of the collaborative exploration.
The objective of this PhD is to design and implement algorithm schemes for the processing and merging of data coming from different sensor types, embedded in several mobile agents in order to obtain an accurate cartography of the environment. The communication and robustness aspects will be addressed as well, in order to coordinate the agents to fulfill the exploration task while ensuring a safe individual behavior.
This work will take place in the wider project of spatial reconstruction of environment and autonomous navigation of a set of mobile agents. The goal is to obtain an accurate reconstruction for a given area. Partial and/or less accurate reconstructions may be built and used by the agents for the autonomous work as well as their navigation.
1. Kubelka, V., Oswald, L., Pomerleau, F., Colas, F., Svoboda, T. and Reinstein, M. (2015), Robust Data Fusion of Multimodal Sensory Information for Mobile Robots. J. Field Robotics, 32 : 447–473.
2. Optimizing computing and energy performances in heterogeneous clusters of CPUs and GPUs.
S. Vialle, S. Contassot-Vivier, T. Jost. Handbook of Energy-Aware and Green Computing, Chapman & Hall/CRC, 2011, 9781466501164
3. Adel Belkadi, Didier Theilliol, Laurent Ciarletta, and Jean-Christophe Ponsart. Robust flocking control design for a fleet of autonomous agents. In 3rd Conference on Control and Fault-Tolerant Systems, SysTol 2016, Barcelone, Spain, September 2016.